Do You Believe in Dog? is approaching our one-year anniversary (Wow! Yay!!!), and in the coming months, we will be opening up the blog to guest posts from other researchers exploring canine behaviour, cognition and welfare.
Give a warm welcome to our first guest,
Clare Browne from the University of Waikato in New Zealand.
Hi Mia and Julie,
(source) |
I would like to claim that everyone is New Zealand is a fantastic dog trainer and we all communicate brilliantly with our dogs, but alas, we’re just like everyone else. It turns out that when people give feedback to dogs during training, we’re often a bit slow. Let me explain...
You’re no doubt aware that if we want to increase the likelihood that a behaviour occurs again, positive reinforcement (AKA “rewarding” -- adding something to keep the behaviour going) will achieve this. The types of positive reinforcement that are most commonly used in everyday dog training are verbal praise, food, and patting/petting. My PhD studies investigated two things: a) how fast are dog owners delivering positive reinforcement to dogs; and b) does it matter if owners are slow in providing dogs with reinforcement?
Not really Clare's gumboots |
To answer the first of these questions, I put on my gumboots and spent many evenings at my friendly local dog clubs, filming owners training their dogs in beginner classes. I collected 1,810 instances where commands were given to dogs. I then went slightly mad and spent months watching videos of people training their dogs. Figure 1 shows how all the dogs responded to their owners, and 44% of the time, dogs did not respond to their owners at all. This one result made me feel like I wasn’t wasting all these years of my PhD – there clearly is a need for research into the efficacy of dog training!
I used some fancy computer software and measured very precisely (down to 25 frames per second) the time between when the owners said the command and when the dogs performed the behavior, like laying down or sitting. I found that owners varied a lot in the time it took them to deliver positive reinforcement to their dogs. Some owners were almost instantaneous with their praise and then the treat followed quickly, whereas others took ages – the longest time was over 6 s! (That might not sound long to you, but try imagining that you’re a Labrador and having to wait 6 s for a treat, all of a sudden it’s a much more serious situation.)
To answer this second question, I had to run an experiment.
I had each dog work inside a large pen, and I sat in the neighbouring pen. A screen in between us prevented the dogs from seeing my body language. I held a pre-training session where the dogs learned that a “beep” sound came just before a feeding device delivered a food treat into their pen. (This “beep” is analogous to us saying, “Good dog”). Once the dogs were comfortable with this, they were taken out of the pen while I set up the equipment. I put two boxes into the pen. These boxes had held dog treats (but were emptied just prior to the experiment) so that they smelled alluring to the dogs. Each box had an open top, and infrared (IR) beams criss-crossed the top surface of both boxes. These IR beams were connected to a computer, so that whenever they were broken (i.e. when a dog sniffed inside a box), the computer delivered a “beep” and then a food treat came from the feeding device.
Click to enlarge image |
There was a catch: when the dogs sniffed inside the ‘non-target’ box, nothing happened. But when they sniffed inside the ‘target’ box, the positive reinforcement food reward was delivered. The dogs, essentially, were being trained to stick their heads inside one of the two boxes (perhaps not necessarily something every owner wants their dog to do, but a good example of a simple behaviour!).
Twenty dogs received positive reinforcement immediately, as soon as they sniffed inside the target box – there was no delay. However, for another 20 dogs, as soon as they broke the IR beams, the computer started a timer and positive reinforcement (“beep” + treat) was delayed by 1 second.
The results were really interesting! Of the 0 second delay dogs, 60% learned the task; but of the 1 second delay dogs, only 25% learned the task – less than half as many with no delay.
Check out this video of a dog learning the task with a 0 s delay:
This suggests that timing of reinforcement is important when training dogs to do something new – so be fast with your praise and treats!!
A question that always gets asked about this research is “Isn’t 1 second a really short time?” Yes, it is a very short time. But if you really watch a dog (I recommend filming them) – see how much, how fast, how far they can move – a lot can happen in 1 second. In real-world, dog training, in 1 second a dog could be across the room, it could have moved out of position, or we could have reached into our pocket.
Finally, I suspect as trainers, we often get away with having sloppy timing because dogs are just so good at attending to our body language. I hypothesise that we are continually giving unintended feedback to dogs – that this feedback may be acting as conditioned reinforcers (much like a body-language-version of the “good dog”). That’s what I suspect anyway ...
Overall, the take home message is be fast with your feedback!
Sending warm regards from Wintry New Zealand,
Clare
Clare with her dogs, Flint & Apple |
Further reading:
Browne C.M., Starkey N.J., Foster T.M. & McEwan J.S. (2013). What dog owners read: A review of best-selling books, Journal of Veterinary Behavior: Clinical Applications and Research, 8 (4) e38. DOI: 10.1016/j.jveb.2013.04.040
Browne C.M., Starkey N.J., Foster M.T. & McEwan J.S. (2011). Timing of reinforcement during dog training, Journal of Veterinary Behavior: Clinical Applications and Research, 6 (1) 58-59. DOI: 10.1016/j.jveb.2010.09.058
More about Clare
© 2013 Do You Believe in Dog?
This comment has been removed by the author.
ReplyDelete